Deep learning for differentiating novel coronavirus pneumonia and influenza pneumonia

Abstract

Background: Chest computed tomography (CT) has been found to have high sensitivity in diagnosing novel coronavirus pneumonia (NCP) at the early stage, giving it an advantage over nucleic acid detection during the current pandemic. In this study, we aimed to develop and validate an integrated deep learning framework on chest CT images for the automatic detection of NCP, focusing particularly on differentiating NCP from influenza pneumonia (IP). Methods: A total of 148 confirmed NCP patients [80 male; median age, 51.5 years; interquartile range (IQR), 42.5-63.0 years] treated in 4 NCP designated hospitals between January 11, 2020 and February 23, 2020 were retrospectively enrolled as a training cohort, along with 194 confirmed IP patients (112 males; median age, 65.0 years; IQR, 55.0-78.0 years) treated in 5 hospitals from May 2015 to February 2020. An external validation set comprising 57 NCP patients and 50 IP patients from 8 hospitals was also enrolled.Two deep learning schemes (the Trinary scheme and the Plain scheme) were developed and compared using receiver operating characteristic (ROC) curves.Results: Of the NCP lesions, 96.6% were >1 cm and 76.8% were of a density <−500 Hu, indicating them to have less consolidation than IP lesions, which had nodules ranging from 5-10 mm. The Trinary scheme accurately distinguished NCP from IP lesions, with an area under the curve (AUC) of 0.93. For patient-level classification in the external validation set, the Trinary scheme outperformed the Plain scheme (AUC: 0.87 vs. 0.71) and achieved human specialist-level performance.Conclusions: Our study has potentially provided an accurate tool on chest CT for early diagnosis of NCP with high transferability and showed high efficiency in differentiating between NCP and IP; these findings could help to reduce misdiagnosis and contain the pandemic transmission.

Artificial intelligence for imaging-based COVID-19 detection: Systematic review comparing added value of AI versus human readers

“…Overall, 771 articles were screened for meeting the selection criteria. A total of 20 studies that met these criteria were assessed for the risk of bias with the QUADAS-2 tool [5,[10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27][28]. Eight studies with a high risk of bias rating in at least two domains were excluded (Fig.…”

“…1, appendix, Table 3). A total of 12 studies were included in our systematic review [5,[16][17][18][20][21][22][23][24][26][27][28]. (See Fig.…”

“…We identified that seven studies used computed tomography (CT) as the main imaging modality [16,[22][23][24][26][27][28], with the remaining five studies focusing on chest X-ray (CXR) imaging [5,[17][18][20][21]. (See Table 1a for details on study characteristics).…”

“…All studies applied a DL model for their AI methodology that included neural networks (Table 1b). For the majority of four studies, China was indicated as the country where the imaging took place [23][24][27][28], with two studies using imaging data from both China and the United States [16,23]. In terms of data source origins, imaging data was considered from various geographical locations, with a focus on patients from the United States for two studies [17][18], two studies from the Netherlands [20,26], one study from Italy [21], and one study from Hong Kong [5].…”

“…with different inclusion criteria for the selection of COVID-19 positive patients and the related characteristics for imaging required, as well as the use of automatic assessment scoring and subgroup analysis. For the reference standard indicated in the included papers, three studies used RT-PCR [20][21]28], three studies applied consensus findings between readers [13,16,18], two studies provided outcome values for both these options [17,19], and in four studies the use of the reference standard was not sufficiently detailed, but included the above (see Table 1a) [5,23,[26][27]. All studies were based on patient selection with the diagnosis of an RT-PCR or nucleic acid amplification (NAAT) test.…”